Coupling Bayesian theory and static acoustic detector data to model bat motion and locate roosts

TL;DR

This paper introduces a novel Bayesian and reaction-diffusion based model to accurately predict bat roost locations by analyzing static acoustic detector data, capturing complex return behaviors.

Contribution

It combines reaction-diffusion theory with domain shrinking to model bat motion, specifically addressing the complex return phase in a novel way.

Findings

Model fits observed tracking data well

Effectively predicts bat roost locations

Captures both dispersion and return phases

Abstract

We propose a novel approach for modelling bat motion dynamics and use it to predict roost locations using data from static acoustic detectors. Specifically, radio tracking studies of Greater Horseshoe bats demonstrate that bat movement can be split into two phases: dispersion and return. Dispersion is easily understood and can be modelled as simple random motion. The return phase is much more complex, as it requires intelligent directed motion and results in all agents returning home in a stereotypical manner. Critically, combining reaction-diffusion theory and domain shrinking we deterministically and stochastically model a ``leap-frogging'' motion, which fits favourably with the observed tracking data.